Seventeen post-January releases clear the size, license and GPQA filters
A broad web scan found 17 distinct model releases reported above 30B total parameters and carrying Apache 2.0, MIT, or modified-MIT terms. Kimi K2.6 leads this permissive-license subset at 90.5% on GPQA Diamond; custom or research-only licenses were excluded.
17
qualifying releases
90.5%
highest reported GPQA
31B–1.6T
total parameter range
3
permissive license families
Reported GPQA scores
GPQA Diamond where identified · %
What stands out
1
GLM supplies three of the strongest entries. GLM-5, 5.1 and 5.2 progress from 86.0 to 91.2 while retaining MIT terms. z.ai
2
Sparse models dominate the largest sizes. DeepSeek V4-Pro reaches 1.6T total parameters, while Kimi K2.6 reaches 1T, with only a fraction active per token. huggingface.co
3
The threshold admits one near-boundary dense model. Gemma 4 31B is reported at 30.7B precise total parameters, just above 30B. huggingface.co
Qualifying releases
| Model | Total params | Release | License | Reported GPQA | Source |
|---|---|---|---|---|---|
| Qwen3.5-397B-A17B | 397B | 2026-02-01 | Apache-2.0 | 88.4, GPQA Diamond | huggingface.co |
| GLM-5 | 744B | 2026-02-11 | MIT | 86.0, GPQA Diamond | huggingface.co |
| MiniMax M2.5 | 229B | 2026-02-12 | Modified MIT | 85.2%, GPQA Diamond | chatforest.com |
| Gemma 4 31B | 30.7B | 2026-04-02 | Apache-2.0 | 84.3%, GPQA Diamond | huggingface.co |
| GLM-5.1 | 754B | 2026-04-07 | MIT | 86.2%, GPQA Diamond | z.ai |
| Qwen3.6-35B-A3B | 35B | 2026-04-16 | Apache-2.0 | 86.0%, GPQA Diamond | huggingface.co |
| Kimi K2.6 | 1,000B | 2026-04-20 | Modified MIT | 90.5%, GPQA Diamond | huggingface.co |
| DeepSeek V4-Pro | 1,600B | 2026-04-24 | MIT | 90.1%, GPQA Diamond | huggingface.co |
| DeepSeek V4-Flash | 284B | 2026-04-24 | MIT | 88.1, GPQA Diamond | fireworks.ai |
| DeepSeek R2 | 671B | 2026-05-07 | MIT | 84%, GPQA | llmcheck.net |
| Darwin-31B-Opus | 31B | 2026-05-14 | Apache-2.0 | 85.9%, reported GPQA | huggingface.co |
| Command A+ | 218B | 2026-05-20 | Apache-2.0 | 76%, reported GPQA | huggingface.co |
| Kimi K2.7 Code | 1,020B | 2026-06-01 | Modified MIT | 89.6%, GPQA Diamond | fireworks.ai |
| Nex-N2-Pro | 397B | 2026-06-02 | Apache-2.0 | 90.7, GPQA Diamond | nerdleveltech.com |
| GLM-5.2 | 753B | 2026-06-16 | MIT | 91.2, GPQA Diamond | z.ai |
| NVIDIA GLM-5.2 NVFP4 | 753B | 2026-06-25 | MIT | 89.39, GPQA Diamond | huggingface.co |
| Hy3 | 295B | 2026-07-06 | Apache-2.0 | 90.4, GPQA Diamond | huggingface.co |
Method: web search across model cards, publisher pages and independent reports, run 22 August 2026. “Parameters” means total parameters, not active parameters; scores are publisher- or source-reported and are not normalized across evaluation harnesses. Exact-date and benchmark-variant gaps are preserved in the table wording.